Auditry: An Agentic AI System for Tax and Financial Auditing

Modern auditing struggles with data volume, regulatory complexity, and manual review limitations.We developed Auditry, an agentic AI system that autonomously executes multi-step audit workflows using FastAPI, Supabase, and React. Drawing on 2024–2025 industry frameworks from OECD, Deloitte, andCapgemini, we demonstrate how agent-based systems differ from traditional automation through goaldirected reasoning and adaptive workflow orchestration. Whilst agentic AI improves scalability (Capgemini reports only 15% of processes are currently autonomous), deployment requires robust governance,explainability, and human oversight. Our architecture addresses concurrent multi-agent operations, regulatory auditability, and data governance specific toregulated financial environments. We present empirical validation through controlled testing scenarios anddemonstrate measurable improvements in audit coverage and detection accuracy whilst maintaining compliance with professional standards